WLAN-Log-Based Superspreader Detection in the COVID-19 Pandemic

Cheng Zhang, Yunze Pan, Yunqi Zhang, Adam C. Champion, Zhaohui Shen, Dong Xuan, Zhiqiang Lin, Ness B. Shroff

TL;DR

This paper introduces a non-invasive framework that uses existing university Wi-Fi logs to detect pandemic "superspreaders" by analyzing network centrality and environmental transmission dynamics. By combining standard concurrent contact data with a novel "asymmetric contact" model, the framework accurately identifies high-risk individuals and simulates targeted quarantine interventions using an SEIR epidemiological model.

What Problem Does This Paper Address?

Effective contact tracing is vital for controlling airborne pandemics like COVID-19, especially in high-density environments like university campuses. Traditional manual tracing suffers from memory bias and delays, while automated client-based apps (utilizing Bluetooth or GPS) face low adoption rates and privacy concerns.

Furthermore, existing infrastructure-based solutions (like passive Wi-Fi tracking) have two major flaws:

  1. Symmetric Bias: They assume transmission only happens when two people are in the same place at the exact same time (symmetric contact), completely ignoring environmental transmission where a virus lingers in a room after an infectious person leaves.
  2. Naïve Metrics: They often rank superspreaders simply by their total number of unique contacts, ignoring how an individual's specific position within a social network topology influences disease propagation.

Methods

The authors propose a multi-step framework that processes raw network data into actionable public health interventions:

  1. Trajectory Reconstruction: Raw WLAN logs are cleaned and sorted to estimate when devices associate and disassociate with specific wireless Access Points (APs), creating a time-series trajectory for each user.
  2. Hybrid Contact Graph Construction: The framework builds two distinct types of contact graphs:
  3. Symmetric Contact: Standard co-location at the same AP for a meaningful duration (e.g., $>15$ minutes).
  4. Asymmetric Contact: Captures environmental infection. If Person A stays at an AP for a long duration (e.g., $>50$ minutes), they heavily shed pathogens. If Person B arrives later or stays briefly (e.g., $5$ minutes), a directed edge is drawn from A to B, denoting directional infection risk. These graphs are merged into a single Hybrid Contact Graph.
Figure 4: Contact tracing using persons’ trajectories. We show trajectories of three persons, i.e., Bob, Alice, and Trudy. At AP2, Bob’s stay time (red) is longer than the environmental infection duration (orange). There is a symmetric contact (blue) between Bob and Alice and an asymmetric contact (green) between Bob and Trudy.
  1. Vertex Centrality Measurement: To find candidate superspreaders, the authors evaluate three structural network metrics on the hybrid graph: Degree Centrality (number of direct ties), Closeness Centrality (topological proximity to all others), and Betweenness Centrality (how often an individual acts as a bridge along shortest paths).
  2. SEIR Simulation & Intervention: An epidemiological Susceptible-Exposed-Infectious-Recovered (SEIR) model simulates a virus outbreak on the network. The framework evaluates the candidates by simulating targeted quarantines (removing the top central nodes) to see which metric best suppresses the outbreak.
Figure 3: Overview of WLAN-log-based superspreader detection. First, we extract contact graphs from WLAN logs via symmetric and asymmetric contact tracing. Second, we perform vertex centrality measurement to discover potential superspreaders. Finally, we simulate adaptive interventions using the SEIR model.

Datasets

The framework was evaluated using a real-world, large-scale WLAN dataset originally collected by Cao et al.

  • Scale: Over 41 million device association/disassociation log entries.
  • Scope: Covers over 5,000 randomized students across 139 days at a large public university campus spanning 1,500 acres and containing more than 200 buildings (equipped with over 8,000 APs).

Key Insights and Ideas

  • Environmental Tracing Matters Most Locally: Incorporating asymmetric contact tracing helps uncover "hidden" superspreaders on daily contact graphs that standard concurrent tracing completely misses.
  • Betweenness is King for Daily Interventions: On a day-to-day basis, individuals with high betweenness centrality are the most dangerous spreaders. This is because they act as bridges connecting otherwise isolated social clusters (e.g., different majors or residential halls). Quarantining them breaks network connectivity.
  • Behavior Stabilizes Over Time: By analyzing student rankings across weeks using the Rank-Biased Overlap (RBO) metric, the authors discovered that campus mobility patterns fluctuate heavily during the first few weeks of a semester but become highly predictable later on. This means administrators can identify long-term superspreaders early in the term.
  • The 20% Quarantine Threshold: When simulating resource-constrained environments, the authors observed a clear turning point: quarantining the top 20% of network-central individuals drastically flattens the curve, but scaling interventions beyond 20% yields diminishing returns.

Key Results

The hybrid graph approach consistently outperformed random quarantines and standard symmetric-only contact tracing across all network metrics.

  • Daily Outbreak Suppression: On single-day contact graphs, using Betweenness Centrality on the Hybrid graph reduced the Total Infected Fraction (T-Inf) to 37.41%, outperforming symmetric-only tracing (42.61%) and no quarantine (48.45%). It also successfully delayed the peak infection time.
  • Long-Term Convergence: On longer-term (weekly) contact graphs, the performance gap between symmetric and asymmetric metrics narrowed. This occurs because dense, repetitive weekly routines cause the top superspreaders in both graphs to highly overlap, showing that early-stage daily tracking is where asymmetric modeling provides the highest value.

Limitations and Future Work

While highly practical, the study has a few limitations:

  • Historical Data: The evaluation relies on campus WLAN logs from 2015. While these accurately reflect normal, unconstrained human mobility patterns, post-pandemic behavior or hybrid learning models might alter network densities.
  • Simplified Asymmetry: The current asymmetric contact model still requires a small window of temporal overlap between users at an AP. The authors note that completely eliminating this constraint to model purely time-delayed environmental transmission (e.g., entering a room hours later) is an important avenue for future work.

Summary updated

Abstract

Identifying "superspreaders" of disease is a pressing concern for society during pandemics such as COVID-19. Superspreaders represent a group of people who have much more social contacts than others. The widespread deployment of WLAN infrastructure enables non-invasive contact tracing via people's ubiquitous mobile devices. This technology offers promise for detecting superspreaders. In this paper, we propose a general framework for WLAN-log-based superspreader detection. In our framework, we first use WLAN logs to construct contact graphs by jointly considering human symmetric and asymmetric interactions. Next, we adopt three vertex centrality measurements over the contact graphs to generate three groups of superspreader candidates. Finally, we leverage SEIR simulation to determine groups of superspreaders among these candidates, who are the most critical individuals for the spread of disease based on the simulation results. We have implemented our framework and evaluate it over a WLAN dataset with 41 million log entries from a large-scale university. Our evaluation shows superspreaders exist on university campuses. They change over the first few weeks of a semester, but stabilize throughout the rest of the term. The data also demonstrate that both symmetric and asymmetric contact tracing can discover superspreaders, but the latter performs better with daily contact graphs. Further, the evaluation shows no consistent differences among three vertex centrality measures for long-term (i.e., weekly) contact graphs, which necessitates the inclusion of SEIR simulation in our framework. We believe our proposed framework and these results may provide timely guidance for public health administrators regarding effective testing, intervention, and vaccination policies.
BibTeX
@misc{zhang2021wlan,
  author = {Cheng Zhang and Yunze Pan and Yunqi Zhang and Adam C. Champion and Zhaohui Shen and Dong Xuan and Zhiqiang Lin and Ness B. Shroff},
  title = {{WLAN-Log-Based Superspreader Detection in the COVID-19 Pandemic}},
  year = {2021},
  eprint = {2102.11171},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2102.11171},
}
Figure